Usage charges

A model call is not charged at a fixed amount just because it returned HTTP 200. It is settled according to the model product's pricing rules:

Token products: first read valid protocol usage. When usage is missing or invalid, a valid non-streaming success JSON object may be estimated from its output, with at least 1 output token; a streaming response must contain model output that the gateway can extract and is marked estimated. See Usage and billing.
Per-image products: charge by the actual number of valid images returned, not directly by the request's n value.
Count Tokens: counts tokens only and does not create a model-call charge.

Before a call starts, available balance after subtracting in-flight request reservations must be greater than zero. After settlement, the account balance may become negative. Each successful charge should appear in the usage log, ledger, and resulting balance.

See Usage and billing for the complete rules.

A model call is not charged at a fixed amount just because it returned HTTP 200. It is settled according to the model product's pricing rules:

Token products: first read valid protocol usage. When usage is missing or invalid, a valid non-streaming success JSON object may be estimated from its output, with at least 1 output token; a streaming response must contain model output that the gateway can extract and is marked estimated. See Usage and billing.
Per-image products: charge by the actual number of valid images returned, not directly by the request's n value.
Count Tokens: counts tokens only and does not create a model-call charge.

See Usage and billing for the complete rules.